My comment on transgender data collection

Comment on Notice of Availability of Proposed Data Collection Standards for Race, Ethnicity, Primary Language, Sex, and Disability Status Required by Section 4302 of the Affordable Care Act (Document ID HHS-OMH-2011-0013-0001)

The deadline to submit your comment is Monday, August 1, 2011!

As a transgender person and a social scientist, I am excited to hear that HHS will be collecting information relating to transgender phenomena. These activities have the potential to bring us valuable information about the prevalence of transgender feelings, thoughts, beliefs and actions in the general population, beyond an often self-selected community that identifies as transgender and participates in the existing surveys. As a social scientist I have some longstanding concerns about the collection and presentation of survey data about transgender individuals, and I hope that your work will improve the situation. Here are some recommendations that I have, for the process of deciding what data to collect and how, and for the data collection itself.

In my experience, many organizations and agencies working with transgender communities repeatedly and consistently make generalizations about transgender populations that are unsupported by any data. For example, the Transgender Law Center found 194 transpeople through unrepresentative “convenience” techniques, of whom 114 reported annual incomes of less than $15,333. A cover article in the San Francisco Bay Guardian summarized it as, “In other words, more than half of local transgender people live in poverty” – an incorrect characterization that was not disputed by the study authors. As any introductory statistics textbook will tell you, prevalence in a convenience sample tells you nothing about prevalence in the general population. No one knows if the sample was representative of “local transgender people.” Presenting it as representative is misleading to the public and can lead to inappropriate funding allocations and badly targeted health initiatives, and possibly even a backlash against transgender people.

I believe that convenience samples can be very useful, for example to show the existence of job discrimination, poverty and prostitution in our community. There is a limit to their usefulness, however, and they are consistently used beyond that limit by social service providers and community advocates. The result is to spread unreliable information, and quite probably to waste taxpayer money and charitable contributions.

Reports like this are often accompanied by a disclaimer; the Bay Guardian article said, “TLC doesn’t claim the study is strictly scientific — all respondents were identified through trans organizations or outreach workers.” Unfortunately, they almost always go on to report the data as if the disclaimer were meaningless: the next sentence reads, “But the data give a fairly good picture of how hard it is for transgender people to find and keep decent jobs, even in the city that is supposed to be most accepting of them.” The reporting of percentages invites this kind of lip service to sampling procedure. Percentages are meaningless in these situations, but they are always reported, and the effect is to dismiss the disclaimer as a formality, encouraging media reporters to do the same.

On your website I see that you anticipate that the Williams Institute and the Fenway Institute will play a strong role in helping you formulate procedures for collecting information on transgender communities. I agree that they have done a lot of good work, and I support their inclusion in any round tables that you convene. However, both institutes have a history of presenting convenience samples as representative. I strongly recommend that you balance their participation with people who are knowledgeable about the appropriate use of sampling.

I am a strong advocate of qualitative research as a means of finding out problems that exist in the world. There are several advocates from the transgender community who have done quality ethnographic and autoethnographic work. One that I know personally is Gail Kramer, who has written the books My Husband Betty and She’s Not the Man I Married under the pseudonym Helen Boyd. I urge you to include in your Roundtables at least one qualitative researcher like Helen.

To my knowledge, only one researcher has done a representative sample of any segment of the transgender community. That is Niklas Långström of the Karolinska Institutet in Sweden. I strongly recommend that Långström, or someone familiar with his survey, be part of your Roundtables. I am also willing to participate, as a transgender person interested in these issues and as a social scientist who has used representative sampling in my professional work.

The curious incident of the healthy transwoman

I’ve noticed that transgender health researchers tend to focus on people with health problems, and that makes sense. Consequently, I’ve often felt a bit guilty talking about transgender health issues. I don’t have a sexually transmitted disease, the worst thing I’m addicted to is sugar, I’ve never been bashed, and I’m not depressed or suicidal. So why should I talk about my health? Why would any researcher want to study someone like me?

The answer comes from Sherlock Holmes, in the story “The Silver Blaze”:

Gregory ( Scotland Yard detective): “Is there any other point to which you
would wish to draw my attention?”
Holmes: “To the curious incident of the dog in the night-time .”
Gregory: “The dog did nothing in the night-time .”
Holmes: “That was the curious incident.”

There’s a fancy word for this: negative evidence. Often, the absence of a salient event can tell you more about the causes of a problem than a hundred events.

I see this all the time in my computer consulting business. If a customer is not getting an image on their computer monitor, it could be caused by a fault in the motherboard, the video card, the video cable, or the monitor. I can turn on the computer and get a blank screen a hundred times, but that doesn’t help me figure out which component is causing the problem.

If I can get a picture even once, however, I can isolate the problem. If I hook the computer up to a different monitor and the display comes on, I know that the monitor is the problem. If I put in a different video card, I know the customer needs a new video card.

This method can work with transgender health as well. We are a diverse group, and there may be something in family background or upbringing that can make the difference between health and sickness.

There are many choices that we make in our lives, and those choices may affect our health. We need to know the consequences of those choices. Even if that knowledge doesn’t ultimately change our decisions, it can prepare us and allow us to plan better.

That is why we need to hear about a whole range of transgender people, not just those that the researchers were able to track down.

The consequences of sampling bias

I wanted to go into a bit more detail about something I’ve mentioned before: that the use of non-representative samples can cause problems down the line. To illustrate this, I want to examine the claims of health disparities that Emilia Dunham lists in her Bay Windows article.

  1. Transgender people take more hormones and have have more surgeries than average.
  2. Transgender people smoke at a 30% prevalence rate, and use other substances to cope with the stress from discrimination.
  3. We’re more likely to suffer from depression and anxiety, and more likely to live with HIV.
  4. 61 – 64% of transgender people have been physically or sexually assaulted.
  5. 41% of transgender people have attempted suicide.
  6. All these percentages skyrocket for transgender people of color and low-income folks.
  7. A startling 1 in 5 transgender people have experienced complete refusal of services from healthcare providers.
  8. If transgender people aren’t referred to with correct names or pronouns or are treated with coldness, they may avoid the office.

Of these statements, only the last one is an existential statement. All the others are statements of prevalence or likelihood that are not generalizable without a representative sample. In my impression, some of them are more likely to be true of the entire transgender population than others. There are chains of causation from transgender actions to these disparities, and the chains are not all the same. Here are some possible causal chains. They are not the only possible ones, but they are the ones that seem likely to me.

First there are the inherent consequences of transgender actions: more hormones and surgery. If you’re only concerned with transpeople who choose to take hormones and undergo surgery, then of course this is true. But if you believe that not all transpeople choose hormones or surgery, and you don’t know how many do, then you have no way of knowing how great these disparities are.

Then there is harassment based on perceptible differences: physical and sexual assault. A lot of this has to do with passing – as one gender or another, not necessarily the one you prefer. The passing does not have to be total: a transperson can avoid a lot of harassment simply by avoiding being noticed. However, note that there is a feedback loop here regarding socioeconomic status: wealthier transpeople can afford higher quality hormones, surgery, hair removal or attachment, clothes, padding, cosmetics and training that can give them (us) a better chance of passing as the target gender.

There is also discrimination based on records or perceptible differences: refusal of healthcare service. There can also be housing, consumer and job discrimination, which can affect some of the factors below.

A transgender person has a number of potential reactions to the harassment or discrimination described above, including: avoidance of healthcare providers, depression, anxiety, substance abuse, suicide attempts. Out of fear of discovery, many transpeople engage in hidden sexual activities, where there is a greater risk of HIV infection.

Completing the vicious cycle I described above are the consequences of poverty, which may in turn result from discrimination: there is greater likelihood of harassment and discrimination (and the consequences that follow from that harassment and discrimination) and sex work (which increases the likelihood of HIV infection).

I know from personal experience, from friends’ anecdotes and from online reading that these disparities do not affect all transgender people. Some people do not choose hormones, some do not choose surgery. Some never take publicly visible transgender actions, and others pass well enough, so they are never harassed or discriminated against. Some are able to deal with the harassment or discrimination they experience without resorting to depression, anxiety or substance abuse, or attempting suicide (which is not a judgment against those who are unable). Some are able to avoid unprotected sex. Some are wealthy enough to avoid the consequences of poverty.

Here’s the problem with sampling: Dunham and other researchers have no way of knowing for sure whether they’ve oversampled from those who choose hormones and/or surgery; those who take publicly visible transgender actions; those who don’t pass enough of the time to avoid harassment or discrimination; those who already have tendencies towards depression, anxiety, substance abuse, suicide or casual sex, for unrelated reasons; and those who have lower incomes. After all, these are precisely the populations that public health researchers are more likely to come into contact with. Without representative samples, they can never prove that these disparities exist to the extent that they claim.

Now I want you to imagine that these researchers actually have been oversampling these higher-risk populations. On one level the consequences are minimal: if these are the populations with the greatest need, then it’s just another way to spend public health dollars on the people who need them the most. But on the image level and the credibility level, there are problems.

I’ve seen on the Web and on television that some people have a stereotype of “tranny” that combines all these factors: a drug-addicted, unpassable, mentally ill hooker with bad plastic surgery. Some people use that stereotype to justify harassment and discrimination against transgender people, and some family members fight against accepting their relative’s transgender feelings because they fear that this will be their fate. These kinds of unsupported survey results feed into those stereotypes.

What if at some point someone does succeed in doing a representative survey, and finds that the drug-addicted, cigarette-smoking sex workers are a small portion of the transgender population, and that the average transgender person is a drug- and disease-free, well-adjusted, successful computer technician making $60,000 a year? What if all the transgender health money was actually better spent on overlapping programs that would serve the needy population just as well? I think someone might feel cheated, and I think there might be a backlash.

There’s also the possibility that we might be missing out on some valuable information. What if we found that there were people who had the exact same background, and the exact same transgender feelings, but one group became drug-addicted HIV-positive hookers and the other became successful computer technicians? We could examine the populations and see what made the difference between health and sickness. It might not be the obvious solution.

This is why we need representative sampling, and this is why you need to comment on the proposal and tell that to Secretary Sebelius.

A critical opportunity in transgender research

The Department of Health and Human Services has just made a big announcement: they will begin collecting data on LGBT issues, including transgender issues. The goal is to document disparities in health care, as well as plain old disparities in health, so that they can be addressed in the future. The plan is to have two roundtables on “gender identity data collection” with “key experts” this summer and fall, and then the “Data Council” will present a strategy next spring.  The department will also collect public comment in various ways, one being through a website called regulations.gov, which is currently down.

If done right, this could be a tremendous help to understanding transgender issues.  “The first step is to make sure we are asking the right questions,” HHS Secretary Kathleen Sebelius told the Washington Post. “Sound data collection takes careful planning to ensure that accurate and actionable data is being recorded.”  As I’ve written before, current research on transgender feelings and actions is severely hampered by the lack of any kind of representative sample.  Just to give you a quick sense, here are ten very basic questions that nobody knows the answer to:

  • How many transgender people are there?
  • How common are the various transgender thoughts, feelings and beliefs?
  • How common are transgender actions like cross-dressing, body modifications, and “soft mods” like shaving?
  • How common are transgender name and pronoun changes?
  • How common are part-time cross-living and full time transition?
  • How often are sexual activities part of transgender activities?
  • How common are diseases and destructive habits among transgender populations?
  • How many transgender people are in long-term relationships?
  • How often are various subgroups targeted by violence and discrimination?
  • How satisfied are transsexuals twenty, thirty or forty years post-transition?

Unfortunately, transgender research is dominated by two camps, the pathologists who make unfounded generalizations based on case studies of their own patients, and the social service providers who make unfounded generalizations based on service recipients, Internet surveys and word of mouth.  Neither of them seem to have understood the idea that while convenience samples can provide the basis for many useful existential statements, prevalence statements based on unrepresentative samples are worthless.

At this point it looks like the roundtables will be heavily influenced by social service providers who only pay lip service to the limitations of their research.  The Plan says, “While HHS is in the beginning stages of developing data collection on gender identity, many researchers (e.g., Williams Institute at the University of California Los Angeles and the Center for Population Research in LGBT Health at the Fenway Institute) have been working on such data collection for several years.”  The Williams Institute produces reports like “Bias in the Workplace” (PDF), an important summary of numerous studies investigating workplace discrimination that repeatedly acknowledges that the studies are based on convenience samples – and then goes ahead and repeats percentage results as though they meant something.

The Fenway Institute employs transgender health advocate Emilia Dunham as a Program Associate, and she also hosted a webinar on the issue.  It seems quite likely that she will be one of the experts at the roundtables.  But in an otherwise solid article for Bay Windows presenting these changes at Health and Human Services, Dunham uncritically repeated several of these unsupported percentages.

There is a very short list of Experts who I think should be on these Roundtables.  The strongest research into transgender issues has been qualitative research: listening, reading and introspection, finding existential statements but not making unsupported claims of prevalence.  I’ve said before that the best qualitative researcher in the transgender community is Helen Boyd, author of My Husband Betty.  At Helen’s recommendation I’ve also read a few things by Raven Kaldera that have been pretty good, particularly his post on female transvestites (which has somehow disappeared from his website).

There is only one person out there who has ever collected data from a representative sample of any transgender community, and that’s Niklas Långström of the Karolinska Institutet in Sweden.  He’s not focused on the transgender community, and he’s associated with the pathologist Kenneth Zucker (who is not someone we want involved), but he does know how to do a national survey, and it would be worth every penny for HHS to fly him over from Stockholm for the Roundtables.

If they can’t get Långström, then I want to be on that Roundtable.  I don’t have a degree in psychology or public health, but I did take an elementary course in statistics, and I learned what you have to do in order to make a generalization.  But what matters more than any qualification is that I care about doing this right.  If they can’t find anyone else who does, I want to be there.

Am I missing anyone? Are you doing quality quantitative research? Please let me know.